Netflix recently launched VOID their subject removal model [under physics laws]
New AI doesn't just erase objects—it simulates gravity and makes them fall realistically.
Netflix has unveiled VOID, a groundbreaking AI model for video editing that performs 'Physics-Aware Deletion.' This moves far beyond simple object removal or background inpainting. Traditional tools erase a subject but leave supported objects—like a vase being held—floating unnaturally in the frame. VOID addresses this core limitation by integrating an understanding of basic physical laws into the generation process. When a supporting element is removed, the model simulates the consequences, such as an object falling due to gravity.
The model's key innovation is its ability to reason about physical constraints. In the example of deleting a person holding a heavy glass vase, VOID doesn't just fill the gap with a static background. It recognizes that the vase has lost its support, calculates its likely weight and trajectory, and generates a new, physically plausible video sequence—in this case, the vase falling and shattering on the floor. This results in edits that are visually coherent and avoid the 'glitch in the Matrix' look common to other AI tools. The research model is publicly accessible on Hugging Face, allowing developers to experiment with its capabilities.
- Performs 'Physics-Aware Deletion,' simulating gravity and support when removing video subjects.
- Generates new, plausible video sequences (e.g., a vase shattering) instead of static background fill.
- Publicly available for testing on Hugging Face under the repository 'netflix/void-model'.
Why It Matters
This significantly raises the bar for believable AI-assisted video editing, moving from simple removal to intelligent, context-aware scene reconstruction.